I would suggest that you try to use the option for setting exp.sys_wide_zip_loads.pw. If you model all loads as constant admittances by using pw = [0 0 1] you will certainly get a solution with YSUM since in this case the network is linear and the solution will be obtained in single iteration. If you get low voltages in the network (say 0.8 pu or lower) it is likely that the operating conditions are not acceptable and such configuration should be discarted. I suspect that this is the problem, you can not mantain constant power requirement if voltages go down and therefore all method divergee.
Whether you can model your loads as constant impedances or constant power is out of the scope of MATPOWER. Depending on the character of consumers in a distribution network probably it is more realistic to expect loads close to constant admittance that to constant power. If you have load static characteristics, i.e. dependence of P and Q on the voltage you may choose appropriate values for the vector pw (not just simply setting [1 0 0] or [0 0 1]). Best regards, Mirko On Mon, 2017-07-03 at 15:56 -0400, Ray Zimmerman wrote: > If you haven’t tried it, certainly try increasing the number of > iterations for the radial methods. > > > Ray > > > On Jul 3, 2017, at 3:15 PM, Andrey Vieira <[email protected]> > > wrote: > > > > for Reconfiguration/Restoration process, ie: > > > > Occurrence of significant concentration of buses/Loads in a given healthy > > region (region that will receive some or all of the disconnected Loads) > > Of the feeder (as shown below) via the relocation of disconnected loads > > (optimization process) due to the insulation of some faulty upstream faults > > Of the off region. > > > > Exemplifying Illustration: > > > > I took the 33bw case as an example in three different load flow execution > > scenarios to exemplify my issue. > > SITUATION A: > > The case 33wb is illustrated below for a specific type of topology. > > For this configuration, there was convergence for the > > 4 methods (Newton, PQSUM, ISUM and YSUM) evaluated. > > > > <pastedImage.png> > > > > > > > > > > > > > > > > > > Situation B: > > Similarly, the case 33wb is illustrated below for another specific type > > of topology. For this new configuration, similar to the previous one, > > there was convergence for the 4 methods (Newton, PQSUM, ISUM > > and YSUM) evaluated. > > <pastedImage.png> > > > > > > > > > > > > > > > > > > Situation C: > > For this new configuration, the 33wb case, shown below, presents a > > particular type of topology in which it has concentrated much load > > on the central feeder. The consequence of this was the non-convergence > > of all 4 methods (Newton, PQSUM, ISUM and YSUM) evaluated. > > <pastedImage.png> > > > > Note: This analysis was also performed for other feeders (case 84, > > case 85, case 135 and case70). It seems to me that in the act of network > > switching (each switching sequence is a possible solution), by > > concentrating Loads in a given region of the network, the possibility of > > non-convergence is high, regardless of the method used. I would like to > > know how to proceed with this problem. For the evolutionary algorithm > > I use needs to evaluate this configuration, even Knowing that such a > > solution is not feasible and possibly will be ruled out by the restriction > > criteria of the optimization that I have adopted. What to do? Increase the > > number of iterations? How to make convergence of power flow occur even for > > those absurd configurations? > >
